Lithium-Ion Battery State Estimation Based on Anode Strain Field Reconstitution Utilizing Optical Frequency Domain...

Area of Science:

• Materials Science
• Electrical Engineering
• Data Science

Background:

• Accurate state of charge (SOC) and state of health (SOH) are critical for battery system performance and safety.
• Traditional monitoring relies on electrical parameters, which can be limited in certain applications.

Purpose of the Study:

• To investigate the use of distributed strain data for real-time prediction of battery SOC and SOH.
• To evaluate the efficacy of optical frequency domain reflectometry and neural networks for battery diagnostics.

Main Methods:

• Employed a phase-sensitive optical frequency domain reflectometer for real-time strain field monitoring in lithium battery anodes.
• Utilized feedforward and long short-term memory recurrent neural networks with strain and strain rate data for SOC and SOH prediction.
• Extracted features like maximum strain and cumulative residual strain for enhanced prediction.

Main Results:

• Distributed strain data achieved 98.3% accuracy for SOC prediction, outperforming single-point measurements (88.8%) and matching electrical parameter predictions (98.5%).
• Long short-term memory networks predicted SOH with 96.3% accuracy using strain data.
• Strain-based predictions achieved high precision without requiring electrical battery data.

Conclusions:

• Purely strain-based data can enable high-precision prediction of battery SOC and SOH.
• Distributed optical measurement offers a robust, non-electrical method for simultaneous monitoring of multiple battery packs.
• This approach supports advanced battery management systems for improved safety and performance.